Roozbeh Dehghannasiri

Single-cell RNA sequencing (scRNA-seq) analyses typically focus on alignment and differential gene expression, being blind to other transcriptome variation. Here we present sc-SPLASH for statistics-first, reference-free discovery on barcoded scRNA-seq and spatial transcriptomics; its independent BKC submodule optimizes barcoded data preprocessing and is approximately 50-fold faster than UMI-tools. sc-SPLASH discovers secreted repeat proteins in immune-like cells of sponge (Spongilla; missing from the reference) and tunicate (Ciona).

Pre-training is a powerful paradigm in machine learning to pass information across models. For example, suppose one has a modest-sized dataset of images of cats and dogs and plans to fit a deep neural network to classify them. With pre-training, we start with a neural network trained on a large corpus of images of not just cats and dogs but hundreds of classes. We fix all network weights except the top layer(s) and fine tune on our dataset. This often results in dramatically better performance than training solely on our dataset. Here, we ask: ‘Can pre-training help the lasso?’. We propose a framework where the lasso is fit on a large dataset and then fine-tuned on a smaller dataset. The latter can be a subset of the original, or have a different but related outcome. This framework has a wide variety of applications, including stratified and multi-response models. In the stratified model setting, lasso pre-training first estimates coefficients common to all groups, then estimates group-specific coefficients during fine-tuning. Under appropriate assumptions, support recovery of the common coefficients is superior to the usual lasso trained on individual groups. This separate identification of common and individual coefficients also aids scientific understanding.

Mouse lemurs (Microcebus spp.) are an emerging primate model organism, but their genetics, cellular and molecular biology remain largely unexplored. In an accompanying paper¹, we performed large-scale single-cell RNA sequencing of 27 organs from mouse lemurs. We identified more than 750 molecular cell types, characterized their transcriptomic profiles and provided insight into primate evolution of cell types. Here we use the generated atlas to characterize mouse lemur genes, physiology, disease and mutations. We uncover thousands of previously unidentified lemur genes and hundreds of thousands of new splice junctions including over 85,000 primate splice junctions missing in mice. We systematically explore the lemur immune system by comparing global expression profiles of key immune genes in health and disease, and by mapping immune cell development, trafficking and activation. We characterize primate-specific and lemur-specific physiology and disease, including molecular features of the immune program, lemur adipocytes and metastatic endometrial cancer that resembles the human malignancy. We present expression patterns of more than 400 primate genes missing in mice, many with similar expression patterns to humans and some implicated in human disease. Finally, we provide an experimental framework for reverse genetic analysis by identifying naturally occurring nonsense mutations in three primate immune genes missing in mice and by analysing their transcriptional phenotypes. This work establishes a foundation for molecular and genetic analyses of mouse lemurs and prioritizes primate genes, isoforms, physiology and disease for future study.

Myriad mechanisms diversify the sequence content of eukaryotic transcripts at both the DNA and RNA levels, leading to profound functional consequences. Examples of this diversity include RNA splicing and V(D)J recombination. Currently, these mechanisms are detected using fragmented bioinformatic tools that require predefining a form of transcript diversification and rely on alignment to an incomplete reference genome, filtering out unaligned sequences, potentially crucial for novel discoveries. Here, we present SPLASH+, significantly advancing biological discovery possible with SPLASH, our recently introduced efficient, reference-free statistical approach. Integrating a micro-assembly and biological interpretation framework, SPLASH+ enables new discoveries including broad and novel examples of transcript diversification in single cells de novo, without the need for cell type metadata, which is impossible with current algorithms. Applied to 10,326 primary human single cells across 19 tissues profiled with SmartSeq2, SPLASH+ discovers a set of splicing and histone regulators with highly conserved intronic regions that are themselves subject to complex splicing regulation. Additionally, it reveals unreported transcript diversity in the heat shock protein HSP90AA1, as well as diversification in centromeric RNA expression, V(D)J recombination, RNA editing, and repeat expansion, all missed by existing methods. SPLASH+ is highly efficient, enabling the discovery of an unprecedented breadth of RNA regulation and diversification in single cells through a new automated paradigm of unbiased transcriptomic analysis.

Spinal Muscular Atrophy (SMA) is caused by a deficiency of the Survival Motor Neuron (SMN) protein. Mutations in SMN disrupt mRNA splicing and translation, leading to maladaptive changes in transcriptomes, proteomes, neuroinflammation, and metabolism, which drive motor neuron degeneration in SMA patients. Using a Drosophila SMA model, we found that systemic depletion of Smn leads to accumulation of RNA:DNA hybrids (R-loops), increased DNA damage, dysregulation of amino acids and sugar metabolism and activation of the innate immune response, recapitulating key pathological features reported in mammalian models and severe SMA patients. Persistent DNA damage in Smn-deficient flies alters cell proliferation rates in larval brains and induces extensive cell death in the developing eye. Importantly here, we show that stimulating the resolution of RNA:DNA hybrids with transgenic human RNAse H1 prevents the accumulation of DNA damage and attenuates the transcriptome and amino acid alterations induced by Smn depletion, mitigating the Smn-dependent cellular and developmental abnormalities, in Smn-deficient flies. Our data suggest that depletion of Smn causes an accumulation of aberrant transcripts and chronic DNA damage, which—along with the altered metabolomic profiles associated with Smn deficiency—trigger systemic inflammatory responses, ultimately affecting neuronal function and survival.

Typical high-throughput single-cell RNA-sequencing (scRNA-seq) analyses are primarily conducted by (pseudo)alignment, through the lens of annotated gene models, and aimed at detecting differential gene expression. This misses diversity generated by other mechanisms that diversify the transcriptome such as splicing and V(D)J recombination, and is blind to sequences missing from imperfect reference genomes. Here, we present sc-SPLASH, a highly efficient pipeline that extends our SPLASH framework for statistics-first, reference-free discovery to barcoded scRNA-seq (10x Chromium) and spatial transcriptomics (10x Visium); we also provide its optimized module for preprocessing and k -mer counting in barcoded data, BKC, as a standalone tool. sc-SPLASH rediscovers known biology including V(D)J recombination and cell-type-specific alternative splicing in human and trans-splicing in tunicate ( Ciona ) and when applied to spatial datasets, detects sequence variation including tumor-specific somatic mutation. In sponge ( Spongilla ) and tunicate ( Ciona ), we uncover secreted repeat proteins expressed in immune-type cells and regulated during development; the sponge genes were absent from the reference assembly. sc-SPLASH provides a powerful alternative tool for exploring transcriptomes that is applicable to the breadth of life’s diversity.

We introduce SPLASH2, a fast, scalable implementation of SPLASH based on an efficient k-mer counting approach for regulated sequence variation detection in massive datasets from a wide range of sequencing technologies and biological contexts. We demonstrate biological discovery by SPLASH2 in single-cell RNA sequencing (RNA-seq) data and in bulk RNA-seq data from the Cancer Cell Line Encyclopedia, including unannotated alternative splicing in cancer transcriptomes and sensitive detection of circular RNA.

Early stages of deadly respiratory diseases including COVID-19 are challenging to elucidate in humans. Here, we define cellular tropism and transcriptomic effects of SARS-CoV-2 virus by productively infecting healthy human lung tissue and using scRNA-seq to reconstruct the transcriptional program in “infection pseudotime” for individual lung cell types. SARS-CoV-2 predominantly infected activated interstitial macrophages (IMs), which can accumulate thousands of viral RNA molecules, taking over 60% of the cell transcriptome and forming dense viral RNA bodies while inducing host profibrotic (TGFB1, SPP1) and inflammatory (early interferon response, CCL2/7/8/13, CXCL10, and IL6/10) programs and destroying host cell architecture. Infected alveolar macrophages (AMs) showed none of these extreme responses. Spike-dependent viral entry into AMs used ACE2 and Sialoadhesin/CD169, whereas IM entry used DC-SIGN/CD209. These results identify activated IMs as a prominent site of viral takeover, the focus of inflammation and fibrosis, and suggest targeting CD209 to prevent early pathology in COVID-19 pneumonia. This approach can be generalized to any human lung infection and to evaluate therapeutics.

The detection of circular RNA molecules (circRNAs) is typically based on short-read RNA sequencing data processed using computational tools. Numerous such tools have been developed, but a systematic comparison with orthogonal validation is missing. Here, we set up a circRNA detection tool benchmarking study, in which 16 tools detected more than 315,000 unique circRNAs in three deeply sequenced human cell types. Next, 1,516 predicted circRNAs were validated using three orthogonal methods. Generally, tool-specific precision is high and similar (median of 98.8%, 96.3% and 95.5% for qPCR, RNase R and amplicon sequencing, respectively) whereas the sensitivity and number of predicted circRNAs (ranging from 1,372 to 58,032) are the most significant differentiators. Of note, precision values are lower when evaluating low-abundance circRNAs. We also show that the tools can be used complementarily to increase detection sensitivity. Finally, we offer recommendations for future circRNA detection and validation.